002 Does Lean Improve Labor Standards Management and Social Performance in The Nike Supply Chain - Distelhorst Et Al. (2016)

This article was downloaded by: [14.139.225.
83] On: 20 April 2016, At: 03:03

Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
INFORMS is located in Maryland, USA
002
Management Science
Publication details, including instructions for authors and subscription information:
http://pubsonline.informs.org
Does Lean Improve Labor Standards? Management and

Social Performance in the Nike Supply Chain
Greg Distelhorst, Jens Hainmueller, Richard M. Locke
To cite this article:

Greg Distelhorst, Jens Hainmueller, Richard M. Locke (2016) Does Lean Improve Labor Standards? Management and Social
Performance in the Nike Supply Chain. Management Science
Published online in Articles in Advance 25 Mar 2016
. http://dx.doi.org/10.1287/mnsc.2015.2369
Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions
This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
approval, unless otherwise noted. For more information, contact permissions@informs.org.
The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
support of claims made of that product, publication, or service.
Copyright © 2016, INFORMS
Please scroll down for article—it is on subsequent pages
INFORMS is the largest professional society in the world for professionals in the fields of operations research, management
science, and analytics.
For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
MANAGEMENT SCIENCE
Articles in Advance, pp. 1–22
ISSN 0025-1909 (print) ISSN 1526-5501 (online) http://dx.doi.org/10.1287/mnsc.2015.2369
© 2016 INFORMS
Does Lean Improve Labor Standards?

Management and Social Performance in
Downloaded from informs.org by [14.139.225.83] on 20 April 2016, at 03:03 . For personal use only, all rights reserved.
the Nike Supply Chain

Greg Distelhorst
Saïd Business School, University of Oxford, Oxford OX1 1HP, United Kingdom, greg.distelhorst@sbs.ox.ac.uk
Jens Hainmueller
Department of Political Science and Graduate School of Business, Stanford University, Stanford, California 94305,
jhain@stanford.edu
Richard M. Locke
Watson Institute for International Studies and Department of Political Science, Brown University,
Providence, Rhode Island 02912, richard_locke@brown.edu
T his study tests the hypothesis that lean manufacturing improves the social performance of manufacturers in
emerging markets. We analyze an intervention by Nike, Inc., to promote the adoption of lean manufacturing in
its apparel supply chain across 11 developing countries. Using difference-in-differences estimates from a panel of
more than 300 factories, we find that lean adoption was associated with a 15 percentage point reduction in
noncompliance with labor standards that primarily reflect factory wage and work hour practices. However, we
find a null effect on factory health and safety standards. This pattern is consistent with a causal mechanism that
links lean to improved social performance through changes in labor relations, rather than improved management
systems. These findings offer evidence that capability-building interventions may reduce social harm in global
supply chains.
Data, as supplemental material, are available at http://dx.doi.org/10.1287/mnsc.2015.2369.
Keywords: corporate social performance; global supply chains; lean manufacturing; human resource management;
labor standards
History: Received January 19, 2014; accepted May 27, 2015, by Bruno Cassiman, business strategy. Published
online in Articles in Advance.
1. Introduction apparel brands1 —have responded by adopting com-

Corporate social performance has become an imperative pliance programs to enforce social standards in their
in strategic management. As stakeholders have grown global supply chains. The goal of these programs is to
increasingly adept at pressuring firms surrounding the improve the social performance of upstream business
partners, primarily in developing countries. They seek
social impacts of their activities (Porter and Kramer
to address stakeholder concerns about the social impact
2006), scholars have linked corporate social performance
of production and to reduce reputational risk for lead
to a variety of positive outcomes, including improved firms (Locke 2013). Yet despite widespread adoption of
access to finance (Cheng et al. 2014), the ability to social compliance programs, research has repeatedly
attract talented employees (Turban and Greening 1997, shown that they yield only limited improvements in
Bhattacharya et al. 2008), increased recommendations social performance (Barrientos and Smith 2007, Egels-
from stock analysts (Luo et al. 2013), and improved risk Zandén 2007, Locke et al. 2007b, Lund-Thomsen et al.
management (Koh et al. 2014). Social performance is 2012). Weak social compliance means that important
particularly important for firms transacting in foreign labor, safety, and environmental standards are violated
jurisdictions where their social license to operate may be in the production of popular consumer goods, placing
subject to the influence of powerful local stakeholders employees’ health and economic well-being at risk.
(Kytle and Ruggie 2005, Henisz et al. 2014).
1
A wide variety of multinational enterprises—includ- The most valuable apparel brands ranked by Millward Brown
ing industry leaders in retail (Walmart, Target, Ikea), Optimor (2013) are Zara, Nike, H&M, Ralph Lauren, Adidas, Uniqlo,
Next, Lululemon, Hugo Boss, and Calvin Klein. All have adopted
electronics (Apple, Microsoft, HP), toys (Mattel, Hasbro), supply chain social compliance programs, which are detailed in their
soft drinks (Coca Cola), and the 10 most valuable corporate citizenship reports.
1
Distelhorst, Hainmueller, and Locke: Does Lean Improve Labor Standards?
2 Management Science, Articles in Advance, pp. 1–22, © 2016 INFORMS
From the managerial perspective, ineffective compli- finding is robust to alternative specifications, including
ance programs threaten corporate social performance controls for divergent labor market trends across coun-
and its associated benefits. Socially irresponsible prac- tries, controls for increased monitoring and enforcement
tices in the supply chain expose lead firms to the risk by Nike, and an examination of pretrends among the
of negative financial shocks associated with activist lean adopters. We estimate a modest effect on health,
campaigns (King and Soule 2007) and the disclosure of safety, and environmental compliance, but it is impre-
socially harmful behavior (Klassen and McLaughlin cisely estimated and more sensitive to specification
1996, Flammer 2013).
choices. This pattern of improvements is consistent

This article explores an alternative approach to with the hypothesis that changes in labor relations
improving social performance in global supply chains. associated with high-involvement work link lean to
We study the relationship between management prac- improved labor standards. We also find heterogeneity
tices and social performance among apparel manu- in workplace improvements by country; although the
facturers in emerging markets. The global apparel intervention significantly raised labor compliance in
industry employs over 25 million in low- to middle- India and Southeast Asian countries, factories in China
income countries (International Labour Organization showed no improvement.
2005) and represents an important entry point for Our work contributes to a deeper understanding
developing countries to global trade in manufactures of how multinational strategy affects social outcomes
(Gereffi 1999). Since 2008, Nike, Inc.—an international in global markets in three ways. First, our findings
leader in the design and retail of athletic apparel, represent the first quantitative evidence linking capa-
footwear, and equipment—has promoted the adoption bility building to improved social performance in
of lean manufacturing in its apparel suppliers. This global production. Although a growing scholarly con-
program provided training in lean manufacturing to sensus affirms the importance of corporate social
supplier management, encouraged the adoption of performance, major questions remain about how to
these management practices, and verified that sup- achieve this performance in global production networks.
plier production lines satisfied a set of lean standards. Capability-building interventions are increasingly pro-
Adoption of this production system required significant moted by both global buyers and external stakeholders
changes to the organization of production, worker to improve supply chain social compliance (Oxfam
participation, and management systems. 2010, Hurst et al. 2011), but empirical evidence on
What are the effects of lean manufacturing on social their impact is limited and ambivalent, leading to
performance? Although the proximal objectives of calls for new empirical work (Lund-Thomsen and
lean are to improve manufacturing performance, we Lindgreen 2014). Our examination of over 300 firms in
hypothesize that replacing traditional manufacturing 11 developing countries and our use of unit fixed effects
practices with lean will also result in improved social for econometric identification allow us to improve
performance. We posit two mechanisms, which may both in terms of internal and external validity on
operate in tandem, that link lean manufacturing to previous research on management practices and social
improved workplace standards. The labor relations compliance, which has relied on small samples and
mechanism holds that increased efforts to motivate and cross-sectional analysis (Locke et al. 2007a).
retain production workers under lean manufacturing Second, this study contributes new understanding
result in improved terms of employment, such as wages about the social effects of modern management in
and benefits. The management systems mechanism globalized production. Various studies have docu-
posits that new managerial capabilities lower the costs mented the effects of lean and related high-performance
of complying with social performance standards. work systems on business outcomes, including pro-
We estimate the effects of lean on social performance ductivity (MacDuffie 1995, Dunlop and Weil 1996,
using panel data from more than 300 factories across Ichniowski et al. 1997), product quality (MacDuffie
11 developing countries between 2009 and 2013. Draw- 1995, Bloom et al. 2013), and financial performance
ing on difference-in-differences estimates, we find that (Huselid 1995). However, research on the social con-
the lean intervention was associated with significant sequences of lean has focused on wages and worker
improvements in factory social performance. Adoption motivation (Appelbaum 2000, Cappelli and Neumark
of lean manufacturing practices led to a 15 percentage 2001, Osterman 2006) or environmental performance
point reduction in noncompliant labor grades.2 This (King and Lenox 2001) in advanced economies. We
instead study lean’s effect on compliance with work-
2
“Noncompliant” grades are those in which auditors detected place social standards, which include wages and ben-
“Serious” or “Critical” violations of labor, health, or environmental
efits but also encompass a broader set of practices
standards, as opposed to “Minor” violations. In the Nike factory audit
scoring rubric (see Table A.4 in the appendix), these are represented
intended to protect employees and local communities.
as C or D grades. More detail on factory social performance data By investigating this relationship among manufacturers
appears in §4. in emerging markets, we contribute to a debate on the
Management Science, Articles in Advance, pp. 1–22, © 2016 INFORMS 3
effects of lean in the developing world. Some research our data and empirical strategy, we present our main
suggests that pressures to adopt lean manufacturing finding: lean adoption produced a substantial reduction
and develop fast turnaround capabilities have led to in poor compliance grades associated with wage and
a deterioration of working conditions in emerging work hours violations. The final section discusses the
market suppliers. Managers lacking the resources to limitations of this study and implications for future
effectively implement modern manufacturing systems research and management practice.
shift the costs of flexible production onto the workforce
in the form of longer hours, lower wages, and more

precarious employment (Dhanarajan 2005, Raworth 2. Managing Social Performance in
and Kidder 2009). However, these claims have yet to Global Production
be subjected to quantitative hypothesis testing. Our Global supply chains link thousands of firms across
findings offer evidence that lean manufacturing can be multiple political and economic boundaries. The diffu-
meaningfully implemented in the context of export sion of global supply chains in an array of different
manufacturing in emerging markets and that its adop- industries—including apparel, electronics, footwear,
tion yields benefits for social performance, linking food, toys, and others—has provided developing coun-
economic upgrading to social upgrading in global tries with needed investment, employment, technology,
supply chains (Barrientos et al. 2011). Most importantly, and access to international markets. At the same time,
because these management practices stand to benefit the social and environmental consequences of this
buyers, suppliers, and workers, lean capability building pattern of economic development have provoked con-
promises greater sustainability than traditional social troversies over the role of global buyers and their
compliance programs. local suppliers, often seen as exploiting low wages
Finally, these findings suggest a strategy for reconcil- and regulatory laxity to produce low-cost goods at
ing tension between the market imperatives of modern the expense of workers’ welfare. As publicized by
supply chain management and social performance. activists and social movements (Harrison and Scorse
Contemporary sourcing strategies such as competitive 2010, King and Pearce 2010), child labor, hazardous
costing, reduced lead times, and smaller order sizes working conditions, excessive working hours, and poor
shift risks onto suppliers and their workforces, thereby wages plague many workplaces in the developing
undermining key goals of social compliance programs world (Verité 2004, Pruett et al. 2005, Connor and
(Dhanarajan 2005, Barrientos 2013, Riisgaard 2009, Dent 2006, Kernaghan 2006). These revelations create
Locke 2013). The global apparel industry, where con- scandal and embarrassment for the global companies
temporary sourcing practices have been argued to be that source from these factories and farms.
particularly deleterious to labor outcomes (Anner et al. In the absence of a strong system of global justice
2012), is a key case for addressing the tension between (Cohen and Sabel 2006) and given the limited ability
sourcing strategy and supplier social outcomes. A major (perhaps willingness) of many national governments
goal of the intervention we study was to improve sup- to enforce their own regulations, an array of actors—
plier capabilities to deal with sourcing trends toward including transnational NGOs (Keck and Sikkink 1998,
smaller orders and more rapid turnaround. We show Seidman 2007), global corporations and industry asso-
that adopting management systems to meet these ciations (Haufler 2001, Bartley 2007, O’Rourke 2003,
demands also led to improved social outcomes. Our Ruggie 2008, Reich 2007), and some developed country
results suggest that contemporary trends in supply governments (Bartley 2007)—began to promote pri-
chain strategy need not depress working conditions vate initiatives aimed at establishing and enforcing
so long as emerging market manufacturers possess labor and environmental standards in global supply
appropriate management capabilities. At the same time, chains. We refer to these initiatives as forms of private
we show that promoting the adoption of new manage- regulation (Vogel 2008, 2010).
ment practices is a major undertaking. In the case we The prevalent model of private regulation involves
examine, it involved securing multiyear commitments establishing supply chain “Codes of Conduct.” In the-
of support from supplier leadership, establishing a ory, these standards are enforced on upstream suppliers
dedicated training facility, and engaging intensively through private audits and the threat of withhold-
with suppliers’ operations personnel. ing orders from noncompliant factories. However, a
In the following section, we introduce supply chain decade of research has demonstrated the limitations of
social compliance programs and the challenges of this strategy for enforcing labor standards. Notwith-
aligning business practice with social performance standing years of effort and significant investments by
goals in global production. We proceed to describe global corporations in developing more comprehensive
Nike’s intervention to promote lean manufacturing, monitoring tools, hiring growing numbers of internal
developing our hypothesis that lean manufacturing will compliance specialists, conducting thousands of fac-
yield improved workplace standards. After describing tory audits, and working with external consultants
and NGOs, working conditions and labor rights have 3. Lean Capability Building in the
improved among some supplier factories but have stag- Nike Supply Chain
nated or even deteriorated in many others (Locke 2013). Facing supply chain challenges in delivery time, prod-
Although pressure generated by antisweatshop cam- uct quality, and workplace conditions, in the late 1990s,
paigns has improved wages in some cases (Harrison Nike began a search for management interventions
and Scorse 2010), the scholarly literature on private for its supply base.3 The Toyota Production System
regulation has generally found persistent noncompli- (Ohno 1988, Womack et al. 1991) was selected for
ance in a variety of workplace standards (Barrientos emulation, and a Toyota consultant was hired to adapt
and Smith 2007, Egels-Zandén 2007, Locke et al. 2007a). lean concepts to footwear manufacturing. In 2002, Nike
Despite private initiatives to equalize minimum work- secured commitments from long-term footwear suppli-
place standards across countries, domestic regulatory ers to implement the lean management and production
institutions and civil society remain key predictors of system it had developed, and a dedicated training
social compliance in global supply chains (Distelhorst center was established in 2004 to train both factory
et al. 2014, Toffel et al. 2015). managers and Nike staff. By May 2011, 80% of Nike’s
One important critique of these programs is that footwear suppliers had committed to adopting the new
they decouple compliance activities from core business system.
practices and thereby limit their impact on supplier Lean concepts have been widely studied and applied
social performance. When needs for external legiti- without a clear consensus on the definition of lean
macy diverge from market demands, firms may design production (Shah and Ward 2007). In this study, we
compliance regimes that conflict with other business
characterize the Nike production system as “lean”
processes, a decoupling that has been observed in other
by reference to common goals and features in lean
corporate ethics regimes (Weaver et al. 1999, MacLean
systems described by key works in the literature. The
and Behnam 2010). Within the global buyers that imple-
features of the Nike system (described in Table 1)
ment compliance programs, sourcing decisions are
included identifying the core value stream and orienting
often decoupled from the enforcement of private regu-
production around this concept; balancing production
lation, resulting in tension between these two functions.
processes using takt time (i.e., the available time for
It is not uncommon to hear complaints from social
production divided by consumer demand); eliminating
compliance managers that their mission is not taken
waste through the reduction of inventory buffers and
seriously by their colleagues in purchasing depart-
works in progress; increasing operator participation in
ments (Harney 2008, p. 213). For their part, suppliers
quality control and problem solving for continuous
complain that, despite lip service paid to ethical com-
improvement; and improving operational stability
pliance, sourcing decisions appear to remain guided
with 5S, standardized work, and visual management
by traditional business considerations, such as price,
techniques (Womack and Jones 1996, MacDuffie 1995,
quality, or turnaround (Ruwanpura and Wrigley 2011).
Shah and Ward 2003).
Some buyers have publicly acknowledged that their
Nike reported business performance gains associ-
own sourcing practices—including the proliferation of
ated with its lean intervention in footwear, includ-
styles, last-minute order changes, poor forecasting, and
ing increased productivity, reduced defect rates, and
overloading supplier capacity—contribute to the very
reduced lead times for both delivery and the introduc-
social performance problems that compliance programs
tion of new models (Nike, Inc. 2012). If these practices
attempt to remediate (Nike, Inc. 2012, Locke 2013).
improved productivity and quality, why did manufac-
In light of the limitations of private regulation, we
turers require outside intervention to adopt them? In
study a supply chain intervention that focuses on
fact, management practices associated with inferior
developing the management capabilities of suppliers,
organizational performance are relatively widespread
rather than enforcing standards through sourcing deci-
even in advanced industrial economies (Bloom and
sions. The immediate goal of capability building is
Van Reenen 2007). The adoption of new management
not to monitor and incentivize socially responsible
practices is not fully explained by superiority in effi-
behavior, but rather to change suppliers’ day-to-day
ciency. It is instead constrained by prevailing intellec-
managerial practices in ways that may also support
tual dispositions, preexisting assumptions about human
improved social performance. Capability building for
behavior, institutional conformity, and asymmetries
social performance has been pursued across a vari-
ety of industries (Locke 2013), but claims of impact
3
have yet to be subjected to quantitative hypothesis The following description of Nike’s lean manufacturing program
for apparel is based on internal documents provided by manage-
testing. The following section describes Nike’s lean
ment and interviews and written correspondence with eight Nike
capability-building initiative and the opportunity it managers over 2011–2014: four directors in lean manufacturing;
provided to test whether such interventions improve one vice president, one director, and one development associate in
social performance in global production. sustainability; and one director in communications.
Table 1 Minimum Definitions of Nike Lean Production System (Apparel Manufacturing)
1. Connect or link at least one process Before lean, almost all apparel factories had physically disconnected sewing, ironing, and packing, with high inventory
to the core value stream buffers between each process. Connecting processes to the core value stream (sewing, in apparel factories) means
physically moving operators and machines into the line, with process cycle time balanced to the line takt time. In
practice, most apparel factories chose to connect ironing and packing at the end of each sewing line.
2. Control inventory via flow racks, Flow racks allow for easy retrieval of inventory on a first-in-first-out basis; kanbans are cards used to signal the start
kanbans, and pull systems and end of production. Both tools support pull systems, which drive production by demand at the end of the
process and reduce waste by eliminating inventory that would ordinarily build up in the value stream to absorb
variability in production processes.

3. Utilize an Andon system to signal The Andon system allows production team members to quickly signal production problems to the entire team.
problems in the line Suppliers must adopt a visual system (e.g., colored flag, card, or digital signboard) to signal problems, such as
production defects, machine malfunctions, or an operator’s need for relief. Depending on the problem, activating
the Andon may temporarily stop production while the problem is addressed.
4. Track appropriate metrics for safety, The minimum definition requires collection of these key performance indicators. Suppliers are expected to use these
quality, delivery, and cost measures to track their performance and drive improvements in the value stream.
5. Use in-station quality inspection The concept of not accepting, making, or passing on a defect is introduced to the line. Operators are asked to
self-inspect their own output rather than depend on end-of-line inspection.
6. Utilize standard work in the core Standardized work involves specifying standards for the rate of production (takt time), required inventory, and
value stream sequence of operator actions. These are written on worksheets located at each work station.
7. Show evidence of 5S and visual 5S (sorting, setting, shining, standardizing, and sustaining) ensures operational stability by eliminating waste from the
management work environment. Sorting removes nonessential tools and materials from the workspace. Setting arranges the
workers, parts, and materials to minimize waste as value-added tasks are performed. Shining maintains the
cleanliness of the workstation and its usability by subsequent operators. Standardizing and sustaining refer to the
institutionalization of these practices. Visual management techniques include signs, shadow boards, tape to mark
walkways and production areas, and colors to indicate performance.
8. Manage the core value stream as a Before lean adoption, each production process was managed by separate supervisors. Once ironing and packing are
single entity rather than individual connected to the end of sewing lines, a single supervisor would be responsible for all processes in that line and the
processes final output.
Note. Nike personnel certified lean production lines in apparel factories by evaluating the adoption of these eight practices.
between visible costs versus hard-to-measure benefits Vietnam and introduced to the Nike lean production
(Guillén 1994, Pfeffer 2007). Implementing modern system. All invitees accepted Nike’s offer to receive
management systems like lean also requires knowledge training and agreed to implement the system in their
that may be difficult to acquire in developing countries own plants. In general, the factories receiving the inter-
(Bloom et al. 2013). For these reasons, supply chain vention were larger plants with preexisting sourcing
capability building is neither unusual nor unique to relationships to Nike.5
Nike. Its lean program resembles well-known initiatives The first group of apparel suppliers committed to
to develop supplier capabilities by Toyota, Honda, and the Nike lean program in 2007 and began meeting
other automakers (Sako 2004). to discuss lean concepts and receive limited training.
The perceived success of the footwear program led A full training curriculum was offered starting in
Nike to expand the lean program to its apparel supply 2009 at the newly opened Apparel Innovation and
chain, which is the subject of our study. The global Training Center in Sri Lanka. The program trained
apparel industry employs tens of millions of workers in supplier managers to oversee the lean transformation of
the developing world (International Labour Organiza- their factories. The training program worked on a self-
tion 2005) and has traditionally offered opportunities for funding model that involved significant commitment
developing countries to integrate with global production from suppliers. Participating factories sent managers to
networks (Gereffi 1999). As of November 2015, Nike the Sri Lanka training center for eight weeks and paid
contracted with 396 apparel factories across 40 countries, tuition to cover program costs.6 The training center
employing over 370,000 workers.4
The first wave of lean adopters came from Nike’s 5
These selection criteria are one reason why cross-sectional com-
Apparel Manufacturing Leadership Forum, a group of
parisons of social compliance outcomes do not produce credible
strategic manufacturing partners with long-term rela- estimates of lean’s effects on social outcomes. Previous research on
tionships to Nike. Subsequent waves of lean-adopters working conditions among Nike’s suppliers found that strategic
were nominated by Nike Apparel Liaison Office direc- partners were more likely to have higher compliance scores (Locke
tors. Senior management from invited suppliers were et al. 2007a). We illustrate the problem with cross-sectional estima-
tions below after reporting the results of our difference-in-differences
initially brought to the footwear training center in
analysis.
6
The lean training program for apparel suppliers lasted 12 weeks,
4
For more detail, see Nike’s global manufacturing map: http:// but managers returned from the Sri Lanka training center to their
manufacturingmap.nikeinc.com (accessed February 21, 2016). home factories during the middle 4 weeks to work on assignments.
was located inside an active apparel supplier, which for a wider range of tasks than in traditional mass
allowed trainees to observe and to practice what they production7 (Appelbaum 2000, MacDuffie 1995). Work-
learned in a lean manufacturing environment. After ers in lean systems integrate quality inspection into
completing the program, trainees worked with a Nike production work, suggest process improvements, and
manager to develop a rollout strategy for their home are more likely to engage in multiple production oper-
factories. They began by establishing a pilot line and ations (Berg et al. 1996, Dunlop and Weil 1996). In the
pursuing one element of the transformation, adding Nike system, workers were trained to conduct in-station
new elements until all were adopted and stabilized.
quality inspection and to communicate problems to

After suppliers completed this reorganization of supervisors and coworkers. They were also trained
production lines, Nike personnel visited the plant to to halt production upon discovering major quality
observe progress and to certify that the lines possessed problems. Workers also became responsible for cleaning
the core elements of lean production. Their minimum and arranging their workstations according to the 5S
definition of “lean” covered eight features, summa- demands. These were significant new responsibilities
rized in Table 1. The production line must connect in comparison to the routinized tasks of traditional
or link at least one process to the core value stream; mass production.
control inventory via flow racks, kanbans, and pull Increased levels of worker involvement may lead to
systems; use an Andon system to signal problems in improved workplace standards through two channels.
the line; track appropriate metrics for safety, quality, First, motivating discretionary effort is key to unlocking
delivery, and cost; use in-station quality inspection; use the performance benefits of high-involvement work
standardized work; show evidence of 5S and visual systems (MacDuffie 1995, Becker and Huselid 1998,
management; and manage the core value stream as a Appelbaum 2000). Whereas individual efficiency under
single entity rather than individual processes. In addi-
traditional mass production can be incentivized through
tion to these criteria, Nike personnel also looked for
piece-rate compensation, in high-involvement work
managerial understanding of these processes and the
systems, “[w]orkers will only contribute their discre-
use of takt time and cycle time to organize production.
tionary effort to problem-solving if they believe that
These practices include key elements of the Toyota
their individual interests are aligned with those of the
Production System (Monden 2012, Womack et al. 1991)
company, and that the company will make a recipro-
and many that appear in studies of modern manufac-
cal investment in their well-being” (MacDuffie 1995,
turing management, including techniques of inventory
p. 201). This may involve raising incentives to reward
control, processes to support quality improvement, and
the collection and analysis of performance indicators performance either individually or collectively (e.g.,
(Bloom et al. 2013, Bloom and Van Reenen 2007). based on the quality or on-time delivery performance
of the entire line). Alternatively, managers may pay
3.1. Hypothesized Mechanisms Linking Lean to an efficiency wage premium to motivate difficult-to-
Social Performance observe dimensions of employee effort (Appelbaum
The intervention described in Table 1 primarily sought 2000, Bailey et al. 2001). Second, high-involvement
to transform the organization of production, rather than work requires increased firm investments in employee
to raise workplace labor, health, and environmental human capital. High-involvement work systems require
standards. Although lean manufacturing emphasizes production workers to acquire skills that were not
the importance of trust and respect in the workplace required in traditional mass production, including
(Monden 2012), Nike’s lean program did not train both technical and interpersonal skills (Cappelli and
suppliers on meeting social standards nor raise social Rogovsky 1994). Firm-led employee training programs
performance demands beyond those applied to other are therefore a key element of high-involvement work
suppliers. Nonetheless, there are theoretical reasons to systems (MacDuffie 1995, Ichniowski et al. 1997, Becker
expect that lean manufacturing may lead to improved and Huselid 1998). Increased need for training in
factory social performance. Drawing on previous high-involvement work systems increases the costs
research we posit two such mechanisms, one stemming
from changes in labor relations and the other from new 7
Traditional apparel manufacturing is commonly referred to as the
management systems. “progressive bundle system,” an application of the principles of
The first hypothesized mechanism involves increased scientific management to clothing production. Garment production is
employee involvement and its effect on labor relations. decomposed into simple operations (e.g., individual seams), and
Lean manufacturing systems, including Nike’s, include workers specialize in performing one operation. To accommodate
elements of high-involvement work: workers possess differences in worker speed, bundles of work-in-progress inventory
buffer each step of production. It yields high levels of labor efficiency
increased skills and knowledge, the opportunity to use
and machine utilization, but large inventory buffers also result in
those skills and knowledge, and the motivation to do long cycle times. A garment that requires just a few minutes of
so (Bailey et al. 2001). Lean involves more decentral- actual labor may take several days to complete the process (Dunlop
ized decision making, giving workers responsibility and Weil 1996, Appelbaum 2000).
of worker turnover; the more employers invest in and nonwage benefits. In theory, workers might be
workers, the more costly it is when workers leave the sensitive to workplace health and safety standards as
firm (Cappelli and Rogovsky 1994, MacDuffie 1995). well. However, we assume that wages and benefits
Thus, managers may improve terms of employment in are generally more influential in determining worker
order to improve employee retention, a major challenge motivation and job satisfaction. Survey evidence from
in many emerging market manufacturers. migrant workers in China finds that they are more than
The key empirical prediction of the labor rela- twice as likely to report “low pay” (80%) as “poor work-
tions mechanism—whether it passes through the need ing conditions” (35%) when reporting why they intend
to motivate discretionary effort or to retain skilled to leave a job (Smyth et al. 2009). On the other hand,
employees—is an increase in wages and nonwage bene- if lean’s effects on social performance are primarily
fits. In addition, other working conditions that influence due to new management systems, we expect to see the
worker motivation and satisfaction may improve as largest effects in technical standards, such as industrial
well, such as total work hours, noise and temperature hygiene, hazardous substances, and emergency egress.8
on the shop floor, and sanitation in worker dormitories. After estimating the effect of lean on labor standards,
Consistent with these predictions, several studies of we shed light on these mechanisms by examining the
U.S. firms show that high-involvement work systems detailed workplace practices associated with Nike’s
are associated with increased employee compensation compliance grades.
(Appelbaum 2000, Bailey et al. 2001, Cappelli and To our knowledge, this is the first study to estimate
Neumark 2001, Osterman 2006). The labor relations the effects of lean on workplace standards across a
mechanism holds that lean will raise labor standards large sample of emerging market manufacturers. Inno-
for similar reasons. vative case study research offered initial support for
An alternative mechanism is that management sys- lean’s effect on labor standards in emerging markets
tems associated with lean manufacturing reduce the (Locke et al. 2007a), but a small sample size raised
marginal cost of complying with certain labor, health, the possibility that these effects were idiosyncratic to
and environmental standards (King and Lenox 2001), particular factories or local labor markets. Moreover,
even if labor relations remain largely unchanged. In other research suggests that the move toward lean
addition to changes to workers’ role in production, production in global supply chains has harmed labor
lean emphasizes the development of process improve- standards in emerging markets. In 2003–2004, Oxfam
ment capabilities (Womack et al. 1991). These modern International led a research project on the supply
management techniques are not widely diffused in chain practices of 20 companies spanning 15 coun-
emerging markets (Bloom and Van Reenen 2007). If tries. On the basis of interviews with factory and farm
noncompliance with certain workplace standards is workers, managers, government officials, union and
the result of flawed management processes, lean may NGO representatives, trading agents, importers, and
provide the tools to correct those processes to ensure staff from major brands and retailers, it concluded
compliance. Examples include the absence of processes that “0 0 0 current sourcing strategies designed to meet
to appropriately label and store hazardous chemicals ‘just-in-time’ delivery (premised on flexibility and fast
or ineffective inventory management that leads to turnaround), combined with the lowering of unit costs,
obstruction of emergency exits. Improved production are significantly contributing to the use of exploita-
planning and reduced cycle time (Dunlop and Weil tive employment practices by suppliers” (Dhanarajan
1996, Appelbaum 2000) may also reduce pressure on 2005, p. 531). According to this study, lean production
worker overtime to meet delivery deadlines (Locke is mimicked rather than genuinely practiced when
et al. 2009). By introducing improved systems of pro- suppliers do not possess the capabilities to cope with
cess improvement, industrial hygiene, and production demands by global buyers for shorter production lead
planning, lean may reduce the costs of remediating times, a greater diversity of products and styles, and
these violations of workplace standards. Consistent lower unit prices. They conclude that “As a result, it is
with this account, previous research on U.S. firms finds most definitely the workers at the labor-intensive stage
that the adoption of lean manufacturing is associated of production who are getting leaned on” (Raworth and
with improved environmental performance (King and Kidder 2009, p. 170). A study by the Clean Clothes Cam-
Lenox 2001). paign of 30 garment factories in Sri Lanka, Bangladesh,
The labor relations and management systems India, and Thailand found that demands by large
hypotheses are not mutually exclusive; both may be at retailers like Walmart, Carrefour, and Tesco for quick
work. However, they offer divergent predictions about turnaround and lower unit costs were undermining the
lean’s effects on factory social performance. If increased
worker involvement necessitates efficiency wages or 8
Both mechanisms predict improvement in overtime compliance.
raises the costs of turnover, we expect improvement in Reduction of excessive overtime may be the result of efforts to please
social performance standards that directly influence employees by reducing the intensity of work or reduced cycle times
employee motivation and well-being, such as wages and improved production planning.
ability of suppliers to comply with codes of conduct third-party audits since June 2012. The purpose of the
(Clean Clothes Campaign 2008). Finally, field research dual system is to allow higher-performing factories
in a footwear factory in China found that lean manu- access to the Nike compliance personnel who can facil-
facturing increased health and safety risks for workers itate improvements beyond the minimum compliance
(Brown and O’Rourke 2007). In light of these conflicting standard. A factory’s progress in lean is not used when
claims, it remains unclear whether lean manufacturing making decisions about audit scheduling or the use of
is part of the problem or part of the solution. third-party versus internal auditors.
Nike divides its factory compliance program into

4. Data and Empirical Strategy two topic areas monitored through two different fac-
Measuring workplace compliance with social standards tory audits: health, safety, and environment (HSE)
is challenging even in advanced economies (Weil 2008). and labor.11 In both, findings are based on auditor
Upstream suppliers in today’s global supply chains are inspection of conditions in factory buildings, inter-
predominantly located in emerging economies, where views with workers and managers, and review of legal
workplace inspectorates may be understaffed, lack documents, timesheets, and wage records. These audits
critical technology for managing data, or both (Interna- summarize factory compliance using a four-point scale:
tional Labour Organization 2011). We address these A (4) to D (1). A description of the scoring rubrics
challenges by measuring factory social performance is reprinted in the appendix (see Table A.4). In labor
with audits that assess compliance with Nike’s supplier compliance, factories that score A or B demonstrate no
code of conduct. These audits use common inspection serious violations of the standards. The key difference
procedures and standards across factories in a range is that A factories have fewer than five uncorrected
of developing countries. This permits for repeated minor issues, and B factories have more than five
observations of compliance with social standards in minor issues to address. In labor compliance, factories
several hundred factories that would otherwise be rated C exhibit at least one “serious” violation of the
difficult for researchers to access.9 code. These include failure to provide basic terms of
Nike evaluates factory compliance with standards in employment, more than 10% of employees working
labor, health, and environmental performance using between 60 and 72 hours each week, and isolated
periodic factory audits. Supplier factories are audited instances of underage labor, verbal harassment, or
for social compliance every 12 to 18 months, according failure to provide minimum legal wage or benefits.
to a schedule that takes into account their previous Factories rated D exhibit “critical” violations of the
compliance ratings and levels of factory risk. Adoption labor code, including denial of auditor access and
of lean manufacturing does not factor into this priority provision of false information, unapproved outsourcing
calculation, and there is no scheduling coordination to other factories, use of forced labor, systemic use of
between the operational teams responsible for lean underage labor, pregnancy testing as a condition of
manufacturing and compliance auditing. employment, failure to accurately record work hours,
One-third to one-half of these audits are conducted and more than 10% of employees exceeding daily work
by Nike compliance personnel, a team of approximately hour limits. The grading rubric for HSE compliance
70 employees. Nike’s in-house auditors have technical follows a similar pattern, with factories rated A or B
expertise in human resource management, engineer- being largely compliant and demonstrating progress,
ing, and health and safety. The remaining audits are
and factories rated C or D exhibiting serious system
performed by third-party auditors. Third-party audit
failures and failing to show improvement. In this study,
vendors are trained by Nike and subjected to annual
noncompliant grades refer to audits resulting in Cs
reviews to ensure that their auditing procedures and
or Ds. We analyze the relationship between these com-
grades align with Nike’s in-house team.10 Factories
pliance grades and individual workplace practices after
that fail to reach a minimum B grade within a defined
presenting our main results.12
timeframe have been required to pay for their own
9
Nike Inc. provided audit results for its apparel suppliers from third-party auditors was 2.35, and that assigned by Nike’s in-house
FY2009 through the first half of FY2014, as well as internal docu- auditors was 2.42, which yields a t-test p-value of 0.74. A chi-squared
ments describing its lean program. Nike also made key managers test of independence fails to detect differences in the compliance
involved in its lean program available for multiple interviews with grade distributions across auditor types (p-value of 0.79).
11
researchers. We agreed to withhold the names of individual suppliers See Nike, Inc. (2014) for full documentation of these standards.
as proprietary information in publications but were not otherwise 12
It is important to note that the specific content of these social
constrained in our presentation of research outcomes. compliance standards varies according to local law. For example,
10
We tested for systematic differences in the labor audits conducted the minimum wage in China is different from the minimum wage
by Nike and third-party auditors. Comparing mean audit scores for in Vietnam. Our empirical approach accounts for this in part by
a subsample of labor audits in the 11 countries studied, we find focusing on within-factory differences in compliance over time.
no significant difference in compliance grades. Using the four- However, a necessary assumption is that these within-factory changes
point grading scale described below, the average grade assigned by in grades are equivalent across factories.
Table 2 Factory Compliance Panel Summary grade. We therefore believe that imputing missing val-
ues is the empirical approach least likely to introduce
Labor HSE
bias, because it retains information from the entire
Imputed values? No Yes No Yes sample of factories in each time period. However, we
Countries 11 11 11 11 also repeat our main analysis with no imputation of
Factories 300 300 332 332 missing data with no change in findings. The larger
Observations 862 2,704 959 2,504
number of imputed values for labor compliance results
Compliance scores (%)
A (4) 1903 1607 001 002

from the larger number of labor scores available early
B (3) 3909 4500 4206 4404 in the panel. Approximately one-half of the factories
C (2) 2600 2403 5306 5109 are located in China, and one-third are in Southeast
D (1) 1408 1400 306 305 Asia. Noncompliant factories (rated C or D) comprise
Observations by country (%) over one-third of our labor panel and one-half of our
Bangladesh 305 208 208 205 HSE panel.
Cambodia 002 004 107 106
China 4704 4303 4301 4209
As of FY2014, factories that adopted lean production
Egypt 006 007 008 009 exhibited better labor and HSE compliance than non-
India 700 606 503 503 adopters. Lean adopters had a mean labor grade of 3.14,
Indonesia 705 807 802 801 compared to 2.63 among nonadopters (p-value < 0.01).
Malaysia 709 807 801 708 For HSE, they averaged 2.79 compared to 2.36 among
Sri Lanka 605 705 500 503
Thailand 1202 1109 1106 1009
nonadopters (p-value < 0.01).13 These differences lend
Turkey 206 304 300 309 initial support to the idea that lean production is
Vietnam 406 508 1003 1009 associated with better social compliance. However,
this cross-sectional comparison cannot rule out the
Notes. Summary statistics for the factory compliance panels in labor and in
health, safety, and environment (HSE), showing pre and postimputation of possibility that lean adopters possess unobserved char-
missing values. Missing values are imputed by using start point imputation, acteristics that explain their higher levels of social
carrying over each factory’s most recent compliance score from preceding performance.
periods. Our panels include only factories with at least two audits over the We use the panel data structure and the gradual intro-
time period. Because Nike uses separate audits for labor and HSE, the samples
duction of lean manufacturing across the supply base
are not identical.
to address concerns about unobserved confounding
We built a panel of factory labor and HSE compliance factors. Table 3 describes the progress of lean adoption
ratings over time (Table 2). These data consist of factory in the factory panels. Although no factory had adopted
audit results from FY2009 to the first half of FY2014. lean at the beginning of the sample period in FY2009,
(The Nike fiscal year starts in June and ends in May.) approximately 20% of factories had implemented lean
Because the lean-adopting factories are all apparel man- before the end of our sample period in FY2014. This
ufacturers, our sample includes only apparel factories progressive proliferation of lean manufacturing allows
us to control for both time-invariant factors associ-
in the same 11 countries as the lean adopters. When
ated with each factory and time-varying compliance
factories are not audited in a given half-year period, we
shocks to the entire pool of factories. We estimate lean’s
impute factories’ compliance scores using the results of
effect on social compliance using a standard two-way
their most recent audit. We consider the imputation
fixed-effects regression model given by
of missing values preferable to the assumption that
data are missing at random. Although we can test Yit = i + t + Leanit + it 0 (1)
whether missingness is correlated with our indicator
for lean adoption, we cannot verify the assumption In this equation, Yit is the compliance score on a
that missingness is uncorrelated with potential social four-point scale from A (4) to D (1), i is a factory
performance outcomes. We also have reason to believe fixed effect that controls for time-invariant unobserved
that labor, health, and environmental conditions in confounding factors, t is a half-year fixed effect to
factories exhibit considerable inertia. These workplace control for common shocks across the pool of factories,
practices are tied to management routines, the local Leanit is our measure of lean adoption, and it is an
labor market, and the priorities of factory leadership. error term with E6 1 1 Lean7 = 0. The parameter of
Empirical evidence suggests that factories’ most recent
compliance scores are informative of their state between 13
Nike compliance grades run on a four-point scale from D (1) to
audits. Analyzing consecutive audits within factories, A (4). We examine compliance scores for apparel factories in Asia,
we find that factories retain identical compliance scores Europe, the Middle East, and Africa as of FY2014 Q1. For labor
compliance, we report audit data for 64 lean adopters and 236
in 73% of consecutive labor audits and 84% of consecu- nonadopters. In HSE compliance, we report 68 adopters and 264
tive HSE audits. Only in 8% of labor audits and 1% nonadopters. p-values are reported from two-sided t-tests assuming
of HSE audits do factories change by more than one unequal variances.
Table 3 Lean Adoption in the Compliance Panel The treatment is also slightly lagged by our coding
of lean adoption according to the state of the factory
Labor sample HSE sample
on the first day of a time period, which ensures that
Year Factories % Factories % lean adoption in our data predates the factory audit.
FY2009 0 0 0 0
Neither our lean measures nor our social compliance
FY2010 12 4 12 4 scores rely upon factory self-reporting, which may be
FY2011 27 9 27 8 subject to biases motivated by self-interest.
FY2012 53 18 56 17
FY2013 60 20 64 19 4.1. Selection Bias

FY2014 64 21 68 20 Because the lean intervention was not randomly
Note. The table displays the count of lean-adopting factories in the panels and assigned to factories, we must consider whether the
their share of all factories at the start of each fiscal year. selection process that led to lean adoption biases our
estimates of . The key concern with selection bias in
interest is , the effect of adopting lean production difference-in-differences models is within-unit varia-
techniques on compliance scores. Under the assumption tion in unobserved characteristics that correlates with
within-unit variation in the treatment. For selection into
of parallel counterfactual trends in the treatment and
the lean program to bias our results—either through
control groups, the average treatment effect on the
Nike’s decision about whom to invite or suppliers’
treated is identified by within-factory change in compli-
decisions whether to participate—factory adoption
ance scores among factories that adopt lean production.
of lean must correlate with changes in unobserved
This strategy improves upon cross-sectional compar-
characteristics that also affect social performance.
isons by eliminating concerns about time-invariant and
As described in §3, factories were invited by Nike
slow-changing differences between lean adopters and
to receive training that ultimately led to adoption
nonadopters, such as factory location, product focus,
of lean manufacturing techniques. Their criteria for
business culture, size, ownership, and levels of man-
selecting invitees included the length of Nike’s business
agerial competence prior to lean adoption. Although
relationship with the supplier, factory size, and the
the four point compliance score is an ordinal variable,
perceived commitment of supplier leadership to engage
we estimate these models using ordinary least squares with Nike’s lean program. These qualities might be
(OLS) rather than ordered probit or logit. The maximum expected to produce higher levels of social performance
likelihood estimator is inconsistent in the presence among the invitees, which limits the credibility of
of fixed effects (Greene 2004), and OLS provides the cross-sectional comparisons between lean adopters
best linear approximation to the conditional expecta- and nonadopters for estimating the effect of lean. Our
tion function (Angrist and Pischke 2008). Below we model is identified by within-factory variation in lean
also conduct robustness checks that use dichotomized adoption over time rather than cross-sectional differ-
versions of the compliance grades. All estimations ences between plants. We investigated the possibility
cluster standard errors at the factory level to account that Nike’s invitation to join the lean program was
for potential serial correlation and heteroscedasticity. correlated with other within-factory changes that might
For robustness we use two measures of lean adoption directly affect factory social performance and confound
at the factory level. The first, lean adoption, is a binary our estimates. Specifically, we researched possible
indicator of whether a factory has any certified lean changes in auditing frequency (were lean adopters
production lines at the start of a given time period. The audited more often?), auditing criteria (did the presence
certification of production lines represents the major of lean lines influence auditors’ evaluations?), and
qualitative shift toward new management systems; it is access to training (did lean adopters receive additional
the result of months of manager training, numerous training on social compliance?). According to a review
changes to production processes, and retraining of of internal documents and interviews with Nike man-
involved workers. Our second measure captures the agement, lean adoption did not produce changes in
intensity of the lean treatment by measuring the share of these aspects of their relationship to suppliers. A sec-
total production lines certified to Nike’s minimum lean ond possible concern is whether factories’ decisions
standard; it varies continuously from 0 to 1. Because to accept Nike’s training invitation correlated with
we use lean certification by Nike staff to measure their ability to demonstrate improvement in social
the intervention, our treatment variables are likely performance, but no suppliers declined the invitation
somewhat lagged. By the time that production lines to receive lean training.
are certified by Nike personnel, supplier factories have We empirically explore threats to inference in three
already undergone an extended process of training ways. We test whether changes in auditing frequency
and production line modifications. Nonetheless, lean confound estimates of lean’s effect. We also examine
certification provides a useful metric because it is whether unobserved labor market trends across coun-
measured against a uniform standard by Nike lean staff. tries may have influenced Nike’s decisions about which
Table 4 Effects of Lean Manufacturing on Compliance Scores
Dependent variable: Labor compliance HSE compliance
Imputed values: Yes No Yes No
Mean score (no lean): 2.58 2.56 2.35 2.33
Model: (1) (2) (3) (4) (5) (6) (7) (8)

Lean adoption 0029∗∗ 0042∗∗ 0008 0005

400115 400145 400075 400095
% lean lines 0057∗∗ 0077∗∗ 0008 0003
400165 400215 400115 400155
Factory FEs Ø Ø Ø Ø Ø Ø Ø Ø
Half-year FEs Ø Ø Ø Ø Ø Ø Ø Ø
Effect magnitudes 4% of dependent variable mean5
Point estimate (%) 11 22 16 30 4 3 2 1
95% CI upper (%) 20 34 27 46 9 12 10 13
95% CI lower (%) 3 10 6 14 −2 −6 −5 −11
Factories 300 300 300 300 332 332 332 332
Total observations 2,704 2,704 862 862 2,504 2,504 959 959
Notes. The table shows OLS panel fixed-effects regression from FY2009 H1 to FY2014 H1. Models are fit using both
start point imputation of missing data and no imputation of missing data. Regression coefficients are shown with
robust standard errors clustered by factory in parentheses. The outcomes are factory compliance grades on a
four-point scale (A = 4, B = 3, C = 2, D = 1) for labor and for health, safety, and environment (HSE). The two codings
of the independent variable are any lean adoption (1 if the factory has adopted any lean lines, 0 otherwise) and
percentage of lean lines (count of lean lines/total lines in factory). Effect magnitudes are expressed in percent of mean
compliance scores for all nonlean factory-half observations. CI, confidence interval.
∗∗
p < 0001.
factories to invite, thereby introducing bias. Finally, we We also test whether these results are driven by our
examine the assumption of parallel trends among lean approach to missing data by comparing estimates from
adopters and nonadopters in the pretreatment period. both imputed (balanced) and nonimputed (imbalanced)
Consistent with the parallel trends assumption, we panel data. Lean’s estimated effect on labor compliance
find no evidence of divergent trends until after lean is larger in the imbalanced panels (Models 3 and 4),
adoption. but statistical comparisons fail to detect significant
differences in effects across alternative approaches to
missing data.
5. Results
One potential concern with the preceding analysis is
Table 4 presents the main results of our estimation
the validity of the parallel trends assumption, which
using two measures of lean adoption. Odd-numbered
implies that average outcomes for lean adopters and
models use the binary indicator of lean adoption, and
nonadopters would follow parallel trends in the absence
even-numbered models use the continuous measure:
of the intervention. To inspect differences in treatment
the percentage of lean-certified production lines in a
and control groups before and after the intervention,
plant. In both specifications, lean manufacturing has a
we estimate a panel model using leads and lags of the
positive effect on labor compliance. The adoption of
treatment, similar to that in Autor (2003). We recode
any lean lines results in an improvement of 0.29 letter
our treatment indicator as the “switch” from the last
grades, 11% of the dependent variable mean (Model 1).
time period of no lean lines to the first time period
Going from zero lean lines to a 100% lean factory is
with any lean lines. We then add binary leads and
associated with an improvement of over half a letter
lags of this indicator to the model. The coefficients on
grade (Model 2). We estimate a small positive effect of
these indicators estimate the differences between lean
lean adoption on HSE compliance on the four-point
adopters and nonadopters at periods just before and
scale, but the coefficients are imprecisely estimated.14
after the adoption of lean in the treatment group.
14
We would have come to apparently erroneous conclusions about 4
X
lean’s effects on HSE compliance if we had relied on estimates Yit = i + t + a leanswitchi4t−a5 + it 0 (2)
using cross-sectional variation. Examining compliance outcomes a=−4
in our sample in FY2014, two-sided t-tests allowing for unequal
variances estimate significant effects of lean adoption on both Our fixed effects remain the same as the ordinary
labor compliance (+0051 grades, = 0013) and HSE compliance panel model. The explanatory variable leanswitchit is a
(+0044 grades, = 0006). binary indicator that takes the value 1 only if factory i
Figure 1 (Color online) Leads and Lags of Lean Adoption variation over time in the effect of the intervention.
The improvement in labor compliance grows consis-
Before lean adoption After lean adoption
Difference in labor compliance
tently in the years following lean adoption, reaching a

0.8
statistically significant level 18 months after adoption.
Two years after certifying their first lean line, lean
plants on average score 0.63 letter grades higher on
grades
0.4
their labor audits than non adopters. In periods before
lean certification, we observe statistically insignificant

positive differences between adopters and nonadopters.
0
These may be the result of the lagged nature of our
treatment indicator. Our coding shows when Nike
formally certified lean production lines, but not when
2 years 1 year Year of 1 year 2 years factories initially adopted lean production practices. In
before before adoption after after
some cases, factories adopted their first pilot lines two
Years before/After lean adoption years before their first lean certification.
Notes. The figure shows the results of estimating Equation (2). Estimated A second concern with the previous tests is that
effects of lean manufacturing on labor compliance in periods before (white we treat the four-point letter grades as continuous
area) and after (gray area) factory adoption of lean manufacturing are shown.
Coefficients are displayed with 95% confidence intervals from panel regression
variables, implicitly assuming that adjacent grades
using four leads and four lags of a lean adoption indicator (robust standard are equidistant from one another. In Table 5 we relax
errors are clustered by factory). The period of adoption represents the first this assumption and replicate the analysis with binary
period in which the factory had any lean production lines on the first day of that transformations of the compliance scores. The first
period. The plot shows no significant differences in labor compliance between
transformation codes A or B as 1, and C or D as 0.
lean adopters and nonadopters before lean adoption, but adopters improve after
switching to lean, with the difference reaching statistical significance 1.5 years Recall that Cs and Ds respectively indicate “serious”
after adoption. Results are based on 300 factories and 2,704 factory-half and “critical” violations of labor standards, as detailed
observations; full regression results are reported in Table A.2 in the appendix. in Table A.4. The second transformation codes only A
as 1 and all other scores as 0. We again use OLS
certifies its first lean production line in period t. The rather than logit/probit because of the inconsistency
four leads and lags of this indicator take the value 1 of the maximum likelihood estimator in fixed-effects
only when factory i certifies its first lean line in the estimation (Greene 2004). Again, we find a significant
time period t − a. For nonadopters, these indicators positive effect of lean adoption on labor compliance,
always take the value 0. The result is a model with nine present in both transformations of the dependent
explanatory variables corresponding to the switching variable. These specifications also highlight that the
period and four leads and lags of that switch. By weak effect on HSE compliance is primarily in moving
estimating coefficients for these leads and lags (a ), this factories up to a B score. The effects reported in the
specification allows us to inspect differences between first two columns of Table 5 are particularly important.
lean-adopting plants and nonadopters before and after As we explain below, the gap between compliant
they certify their first lean lines. If unmodeled differ- (A and B) and noncompliant (C and D) grades is largely
ences in compliance trends between the treatment and a function of accurate payment of worker wages and
control groups are driving our results, we may observe benefits (see Figure 2). We estimate in column (1) that
differences in labor compliance between adopters and lean adoption reduces the probability of receiving a
nonadopters before the introduction of lean manu- noncompliant grade by 15 percentage points.
facturing. Apart from providing this check of the Finally, we examine the possibility that improvements
parallel trends assumption, this model also allows us might be explained by increased auditing of the lean
to examine how the effect of lean emerges over time. plants. Because audits identify noncompliant practices
Figure 1 plots the estimated coefficients for these that factories are instructed to improve, they might
indicators, including 95% confidence intervals, high- stimulate improvements in working conditions. If lean
lighting the periods before and after lean adoption. We adopters received more frequent compliance audits,
detect no significant placebo effects in the two years their labor standards could improve because of the
before lean adoption, which suggests that unmodeled effect of auditing. We reestimate our models with
differences between adopters and nonadopters did not indicators of the cumulative number of audits each
significantly affect labor compliance and therefore the factory has received and find no significant difference
parallel trends assumption seems plausible. We would in the estimated effect of lean adoption (see Table A.1 in
not expect to see parallel pretreatment trends if lean the appendix). This is consistent with previous research
adopters were cherry-picked based on recent improve- showing limited improvements in labor standards from
ments in social performance. The figure also illustrates repeated factory inspections (Locke 2013).
Table 5 Binary Transformations of Dependent Variables
Dependent variable (DV): Labor compliance HSE compliance
DV coding AB = 1, CD = 0 A = 1, BCD = 0 AB = 1, CD = 0 A = 1, BCD = 0
DV mean (no lean): 0.59 0.14 0.39 0.003
Model: (1) (2) (3) (4) (5) (6) (7) (8)

Lean adoption 0015∗ 0008 0010 0000

400075 400055 400065 400005
% lean lines 0028∗∗ 0024∗∗ 0012 0001
400115 400085 400105 400015
Factory FEs Ø Ø Ø Ø Ø Ø Ø Ø
Half-year FEs Ø Ø Ø Ø Ø Ø Ø Ø
Factories 300 300 300 300 332 332 332 332
Total observations 2,704 2,704 2,704 2,704 2,504 2,504 2,504 2,504
Notes. The table shows OLS panel fixed-effects regression from FY2009 H1 to FY2014 H1 (11 periods). Regression coefficients are shown with robust standard
errors clustered by factory in parentheses. The outcomes are binary transformations of factory compliance scores for labor and for health, safety, and environment
(HSE). The first transformation codes factories receiving A or B grades as 1, and 0 otherwise. The second transformation codes factories receiving only an A grade
as 1, and 0 otherwise. The two codings of the independent variable are lean adoption (1 if the factory has adopted any lean lines, 0 otherwise) and percentage of
lean lines (count of lean lines/total lines in factory).
∗
p < 0005; ∗∗ p < 0001.
5.1. Mechanisms for 442 audits.15 We estimate the importance of 27 audit

Above we identified two mechanisms that may drive line items for predicting overall labor compliance scores
these results. The labor relations hypothesis held that using the random forest algorithm (Breiman 2001).
increased demands on worker skill and discretionary Random forests are among the most popular techniques
effort incentivized managers to improve the terms of to emerge from the machine learning literature, in part
employment. In contrast, the management systems because they offer extreme flexibility (Varian 2014).
hypothesis focused on how lean changes the marginal They do not require parametric assumptions on the
costs of compliance from the managers’ perspective. By functional form relating predictors to outcomes and
developing capabilities surrounding process improve- can accommodate a range of nonlinearities and interac-
ment, industrial hygiene, and production planning, tions that are difficult for researchers to specify when
lean may reduce the costs of complying with technical faced with a large number of predictors. In evaluating
workplace standards. The labor relations hypothesis the relative contribution of over two dozen practices
predicts improvement in terms of employment that to the overall grade, we faced exactly this modeling
directly bear on workers’ motivation and job satisfac- challenge. We separately conducted analyses of the
tion, such as wages and benefits. The management four-point compliance grade (A–D) and a dichotomized
systems hypothesis predicts improvement in technical version: compliant grades (As and Bs) versus non-
areas of compliance, such as hazardous materials and compliant grades (Cs and Ds). A similar analysis of
emergency egress. HSE compliance grades appears in Figure A.1 in the
Lean’s effect is predominantly on labor standards
appendix.
rather than health, safety, and the environment (Table 4).
Consistent with the labor relations hypothesis, these
The difference between point estimates for labor and
results show that variation in labor compliance scores
HSE is statistically significant for the continuous mea-
reflects important differences in employee compensa-
sure of lean, but not the binary indicator of lean
tion. Figure 2 plots variable importance scores for these
adoption. Because Nike’s labor audit includes mea-
workplace outcomes, which are grouped by category.16
sures of wages and benefits, this offers preliminary
The most important workplace practices for predicting
support for the labor relations hypothesis. However,
this depends on whether labor compliance grades are
actually informative of terms of employment that are 15
Ideally we would estimate the effects of lean on each workplace
important to worker motivation and satisfaction. To practice, but our data include just 16 prelean and 34 postlean
shed light on this question, we analyze the relationship observations of the treatment group.
16
between labor compliance grades and detailed work- Variable importance scores summarize the predictive power of
variables in classification trees, which do not yield coefficients in
place practices. We merged compliance grades with
the same way as traditional regression models. Importance scores
records of detailed workplace practices for a subset measure the total decrease in node impurities from splitting on a
of audits. This allows us to analyze the relationship given variable, averaged over all trees. Node impurity is measured
between compliance grades and workplace practices by the Gini index (Liaw and Wiener 2002).
Figure 2 (Color online) Predictors of Labor Compliance Grades
4 Point A–D AB versus CD

Total compensation:
Wages accurately calculated and paid
One day off per seven days of work
Legally mandated benefits provided
Time-keeping system for work hours
Work hours under maximum

Leaves for illness, vacation, maternity
Payroll docs complete and accessible
Benefits payments deposited in accounts
Minimum wage compliance
No disciplinary fines
No penalty for declining to work overtime
No forcing overtime work
Hiring practices:
Trains employees on terms of employment
Employment records for all employees
Non-discrimination compliance
Employees free to terminate employment
No underage employees
Does not assign home work
Employees communication:
Confidential grievance system
Procedures to investigate grievances
Employees believe grievance system is fair
Employees can freely associate
No factory interference with employee orgs
No coaching worker responses to auditor
Treatment of employees:
No abusive treatment of employees
Leave for emergencies or medical care
Access to drinking water and toilets
0 5 10 15 20 0 5 10 15 20
Variable importance from random forest
(mean decrease in Gini index)
Notes. The figure shows random forest estimates of variable importance from sample of audits with detailed subscores for labor compliance, with a total of
442 audits over FY2008–FY2012. Variable importance is measured by the sum of all decreases in the Gini impurity index associated with splitting on a given
variable, normalized by the number of trees in the forest (Breiman 2002, Liaw and Wiener 2002).
labor compliance grades are related to employee com- 5.2. Heterogeneous Effects by Country
pensation and hours, especially accurate payment of Finally, we examine the effects of lean in different
wages, one day off per seven days of work, provision of countries. We interact the lean measures in our panel
legally mandated benefits, time-keeping for work hours, model with country indicators to estimate country-
and keeping work hours under the maximum limit. specific treatment effects. The seven countries that
In contrast, HSE grades, where we find no significant occupy at least 5% of the sample each have their own
indicators, and the remaining countries are pooled into
improvement, are primarily determined by technical
a residual indicator comprising Bangladesh, Cambodia,
and procedural standards that do not play a major role
Egypt, Indonesia, and Turkey.
in worker motivation and retention. The top predictors We find significant heterogeneity in the treatment
are risks from confined spaces and the management effect across countries (Figure 3). In India, Malaysia,
of hazardous substances (see Figure A.1). Although and Thailand, any lean adoption is associated with
these are important working conditions, workers cite improvement of over half a letter grade in labor com-
them significantly less often than wages as a reason for pliance. The effect in Vietnam is smaller but statistically
leaving the enterprise (Smyth et al. 2009). significant. However, in China, Sri Lanka, and our pool
Figure 3 (Color online) Country-Specific Treatment Effects industry, we estimate that adoption of any lean manu-
facturing is associated with an improvement of 0.29
India labor compliance grades, or a 15 percentage point
increase in the probability of compliance (A or B grades).
Thailand
Although this finding by no means obviates concerns
Malaysia about working conditions in emerging markets, it does
provide the first quantitative evidence on the efficacy
Vietnam
of capability building in improving labor standards in

Other global supply chains.
Sri Lanka
These results are consistent with the hypothesis that
Lean adoption lean changes labor relations in ways that stimulate
China % lean improved workplace standards. Because lean requires
increased investments in worker training and higher
–1.0 –0.5 0 0.5 1.0 1.5 2.0 2.5
levels of discretionary effort, managers have an incen-
Effect of lean on labor compliance tive to improve labor conditions in order to retain and
Notes. The figure displays point estimates with 95% confidence intervals motivate skilled employees. We find that the top predic-
(using robust standard errors clustered by factory) from a fixed-effects model tor of labor compliance grades is wage compliance, an
interacting country indicators with treatment variables. The two codings of the
independent variable are lean adoption (1 if the factory has adopted any lean
improvement that is difficult to explain by improved
lines, 0 otherwise) and percentage of lean lines (count of lean lines/total lines in techniques of process improvement, production plan-
factory). Countries pooled in the “other” indicator are Bangladesh, Cambodia, ning, or industrial hygiene. At the same time, these two
Egypt, Indonesia, and Turkey. Results are based on 300 factories and 2,704 mechanisms are not mutually exclusive, and we can-
factory-half observations. Regression results are available in Table A.3 in the not exclude the hypothesis that management systems
appendix.
also play a role. It is also important to consider that
compliance with some health, safety, and environment
of residual countries, lean adopters do not improve standards may require structural changes to the plant,
significantly. F -tests reject the hypotheses that the effect such as constructing new fire exits, changing wiring,
in China is identical to those of India (p-value < 0.01), or investing in pollution mitigation equipment. Such
Thailand (p-value < 0.01), and Malaysia (p-value < 0.03). changes may be more costly and slow changing than
Although lean adoption appears to have a large effect adopting new practices in worker hiring, compensation,
on labor standards in several key apparel-exporting and hours.
countries, we detect no effect on factories in China, Future quantitative and qualitative research may
where nearly half of our sample is located. clarify the relative contributions of the labor relations
This country-level heterogeneity in effects raises and management systems mechanisms. One empirical
the possibility that country-specific labor market or approach would involve estimating lean’s effect on
regulatory trends may be confounding our estimates of more narrowly defined workplace outcomes whose
the effects of lean. Factories in countries experiencing improvements are attributable to one mechanism but
wage increases or more stringent regulations may be not the other. Taking Nike’s compliance subscores as
more motivated to embrace process changes to maintain an example, improvements in wages, verbal abuse by
competitiveness. At the same time, those within-country supervisors, disciplinary fines, and worker grievance
labor market trends may exert their own effects on systems seem unlikely to be explained entirely by
workplace standards. To address this possibility, we improved management sytems. On the other hand,
estimate models that allow for country-specific linear standards dealing with wastewater management and
and quadratic time trends. These controls model the fire safety precautions appear unlikely to play a major
possibility that countries exhibit divergent trends in role in worker motivation and retention. Studying
workplace compliance unrelated to the presence of lean the effects of lean on these fine-grained workplace
manufacturing. The estimated effects of lean attenuate outcomes, including worker turnover rates, could
slightly but remain statistically indistinguishable from shed light on causal mechanisms. Qualitative research
the main estimations (see Table A.1). on manufacturers negotiating changes in production
systems would also be highly informative. Key interme-
diate outcomes—such as new kinds of worker training
6. Discussion and new channels of communication between workers
Our findings support the hypothesis that lean manufac- and management under the labor relations hypothesis—
turing and associated high-involvement work practices may be obvious from focused examination of individual
can improve social performance in emerging mar- workplaces but challenging to quantify for a large
ket manufacturers. Based on a five-year analysis of a sample of factories in the absence of panel surveys of
capability-building intervention in the global apparel employees and managers.
The effect of lean on labor standards was strongest through the threat of external sanction. Buyers mandate
in India, Malaysia, Thailand, and Vietnam. However, that suppliers meet social responsibility standards in
we detect no effect of the lean intervention in Sri order to do business. For the sanction-based system
Lanka, China, and the pool of remaining countries. It is to work, the buyer has to be willing to bear the costs
perhaps unsurprising to find little improvement in Sri of adequately financing an auditing team to monitor
Lanka, a country known for high levels of factory social compliance as well as switching costs associated with
compliance (Ruwanpura and Wrigley 2011). Among terminating relationships with noncompliant suppliers.
factories in Sri Lanka, 85% of nonlean observations In turn, the supplier must believe that investments in
exhibited a labor compliance rating of B or higher, improved conditions are more valuable than losing
with 31% receiving A ratings. The same cannot be said the buyer’s business. However, the last decade of
for China, where just 57% of nonlean observations research offers evidence that these programs offer lim-
received a B or higher, and only 8% received an A grade. ited improvements, because buyers continue business
China’s labor compliance grades are comparable to relationships even under conditions of sustained non-
factories in India (48% B or higher, 10% As), where compliance (Barrientos and Smith 2007, Egels-Zandén
factories exhibited significant improvement. 2007, Lund-Thomsen et al. 2012, Locke 2013, Distelhorst
One clue to the absence of an effect in China is et al. 2014). Even when traditional compliance regimes
the lower intensity of lean adoption. By the start of function as designed, the buyer must continuously
FY2014, all 12 Thai lean adopters, 9 of 10 in Malaysia, apply these pressures, with their associated costs for all
and all 3 in India had certified more than 33% of their parties, to sustain improved workplace conditions. In
production lines to meet Nike’s minimum definition. contrast, previous research shows that lean and other
In contrast, 6 of the 16 lean adopters in China had less forms of modern manufacturing deliver substantial
than 33% lean lines. If the effects of lean are associated benefits to business performance (Ichniowski et al. 1996,
with a certain threshold level of adoption, then perhaps Bloom et al. 2013). If these practices can be successfully
factories in China have not yet reached that threshold. introduced, suppliers themselves may have a stake in
A second possibility is that features of China’s indus- maintaining them.
trial workplace create barriers to the development of How generalizable are our findings? The interven-
high-involvement work systems that devolve mean- tion we studied was firm driven and implemented in
ingful decisions to workers. Previous research on lean the naturalistic setting of actual manufacturers across
and high-performance work systems emphasizes the 11 emerging markets. The manufacturing practices
importance of complementary “bundles” of work and that comprise Nike’s lean intervention (Table 1) are
personnel practices in delivering benefits for the firm widely known and commonly employed in a range of
(Milgrom and Roberts 1990, MacDuffie 1995, Dunlop manufacturing settings. Because these practices are
and Weil 1996). If changes in labor relations are the not highly idiosyncratic or proprietary, our findings
primary mechanism through which lean affects labor may be applicable to similar interventions by other
standards, key elements of the bundle may include insti- firms. Our findings also come from an industry in
tutions that facilitate communication between workers which emerging markets play an important role. After
and management and foster the necessary trust to China and the European Union, the world’s biggest
give workers meaningful decision-making authority. exporters of apparel are Bangladesh, Vietnam, India,
Contemporary China has highly limited institutions and Turkey (World Trade Organization 2014). With low
of worker voice, with the official labor union failing barriers to entry, apparel manufacturing is viewed as a
to provide significant bottom-up representation of “starter” industry for growth strategies that emphasize
worker interests and opposing higher levels of worker export-oriented industrialization (Gereffi 1999). In these
participation in decision-making (Friedman and Lee ways, our findings on lean manufacturing and labor
2010, Brown and O’Rourke 2007). In the absence of standards are theoretically applicable to firms across a
institutions facilitating voice and trust, employers may variety of emerging markets.
adopt the physical and managerial elements of lean At the same time, there are important limits to general-
without implementing the high-involvement work izability. As a model for improving social compliance in
practices that stimulate improved labor standards. global supply chains, capability-building interventions
Learning more about mechanisms and heterogeneous may be limited to large buyers like Nike. Large multi-
effects across workplaces is the focus of our future nationals have more resources to support training
research. Regardless of the mechanisms that produce programs, and the scale of their orders makes it eas-
positive spillovers for social performance, lean capa- ier to persuade suppliers to invest in implementing
bility building differs in fundamental ways from the new management systems. This intervention also tar-
traditional private regulatory approach to social respon- geted suppliers with long-term business relationships
sibility in supply chains. The dominant mode of private with Nike. If such relationships are a precondition for
regulation attempts to improve workplace conditions intensive collaboration on management systems, this
intervention may not be plausible in supply chains attempt to create value for both the buyer and sup-
with high supplier turnover. In short, we should nei- plier, such that both parties have incentive to cultivate
ther over- nor undergeneralize from the results of and sustain new management practices. By offering
this research. Lean manufacturing is associated with evidence that certain forms of capability building
improved labor standards in this important case, but it enhance factory social performance, we identify a
would be prudent to replicate these analyses with new specific opportunity to create “shared value” in global
industries and lead firms. supply chains (Porter and Kramer 2006). If global
buyers, supplier management, and the production
One important limitation of this study is our limited

data on the features of individual factories and their workforce simultaneously derive benefit from this
sourcing relationships to Nike. It may be instructive in approach to manufacturing, lean capability building
future research to explore the effects of lean manufac- may represent a form of self-enforcing institutional
turing on buyer order volume and, more generally, on change that supports improved working conditions in
the relationship between buyers and suppliers. Capa- emerging markets.
bility building programs may produce higher levels of Supplemental Material
trust and relational contracting in the buyer–supplier Supplemental material to this paper is available at http://dx
relationship. A common complaint from developing .doi.org/10.1287/mnsc.2015.2369.
world suppliers is that buyers demand improvements
in factory labor conditions but lack commitment to Acknowledgments
The authors thank Lea Borkenhagen, Steve Castellanos, Hang-
continued sourcing (Ruwanpura and Wrigley 2011,
Chih Chang, Angela Ho, Sharla Settlemier, Scoti Snider,
Locke et al. 2009). Suppliers’ belief that customer rela- Stephani Kobayashi Stevenson, and Tom Young at Nike Inc.
tionships are fragile and short-lived reduces incentives for their support of this research. They also thank Joshua
to invest in social compliance. In the Nike case, capa- Cohen, Eli Friedman, Retsef Levi, Anita McGahan, Will
bility building targeted suppliers that already enjoyed Mitchell, Charles Sabel, and Mari Sako, as well as seminar
long-term sourcing relationships with Nike. However, participants at the Massachusetts Institute of Technology,
in other cases, capability building may credibly signal University of Oxford Saïd Business School, the Wharton
commitment to an extended sourcing relationship, School, the New School for Social Research, University of
Toronto, the Labor and Employment Relations Association,
thereby increasing supplier trust that investments in
and the American Political Science Association for helpful
social performance will not go to waste. feedback. Funding for this project came from the MIT Sloan
Our findings have straightforward implications for School of Management Dean’s Innovation Fund and a grant
multinational management practice. Capability build- from CREATe.org (Center for Responsible Enterprise and
ing diverges from traditional private regulation in its Trade). Any remaining errors are the authors’ own.
Appendix
Table A.1 Effects of Lean, Controlling for Cumulative Compliance Audits and Country Trends
Dependent varaible: Labor compliance HSE compliance
Model: (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
∗∗ ∗∗
Lean adoption 0029 0029 0021 0008 0004 0000
400115 400115 400115 400075 400075 400065
% lean lines 0057∗∗ 0057∗∗ 0044∗∗ 0008 0000 0004
400165 400165 400155 400115 400115 400105
Cumulative audits 0010 0004 0010 0004
400135 400135 400135 400135
(Squared) −0001 0001 −0001 0001
400035 400035 400035 400035
Country time trends Ø Ø Ø Ø
Factory FEs Ø Ø Ø Ø Ø Ø Ø Ø Ø Ø Ø Ø
Half-year FEs Ø Ø Ø Ø Ø Ø Ø Ø Ø Ø Ø Ø
Factories 300 300 300 300 300 300 332 332 332 332 332 332
Total observations 2,704 2,704 2,704 2,704 2,704 2,704 2,504 2,504 2,504 2,504 2,504 2,504
Notes. Alternative specifications of the compliance models estimated in Table 4 are shown, controlling for cumulative audits and country-specific time trends (linear
and quadratic) for each of the 11 countries in the sample. Cumulative audits are the total number of audits the factory has experienced as of the current period. The
outcomes are factory compliance grades on a four-point scale (A = 4, B = 3, C = 2, D = 1) for labor and for health, safety, and environment (HSE). The two
codings of the independent variable are any lean adoption (1 if the factory has adopted any lean lines, 0 otherwise) and percentage of lean lines (count of lean
lines/total lines in factory).
∗∗
p < 0001.
Table A.2 Effects of Leads and Lags of Lean Adoption on Labor Compliance
Dependent varaible: Labor compliance
Model: (1)
leanswitch t+4 000272

40006585
40009665
4001115
4001265
leanswitch t 00115
4001635
leanswitch t−1 00220
4001805
leanswitch t−2 00283
4001935
leanswitch t−3 00490∗∗
4002055
leanswitch t−4 00548∗∗
4002165
Factory FEs Ø
Half-year FEs Ø
Factories 300
Total observations 2,704
Notes. The table shows OLS panel fixed effects regression from FY2009 H1 to FY2014 H1.
Regression coefficients are shown with robust standard errors clustered by factory in parentheses.
The outcomes are factory labor compliance grades on a four-point scale (A = 4, B = 3, C = 2,
D = 1). The binary indicator leanswitch t takes the value 1 only in the first period after lean
adoption. The leads and lags of this indicator allow us to examine differences between the
treatment and control groups before (t + a) and after (t − a) lean adoption. The results are plotted
in Figure 1.
∗
p < 0005; ∗∗ p < 0001.
Table A.3 Country-Specific Effects of Lean on Labor Compliance
(1) (2)
Lean adoption
× China −00074
4002255
× Thailand 00704∗∗
4001825
× Vietnam 00239
4002115
× Sri Lanka −00023
4002405
× Malaysia 00608∗∗
4002205
× India 00809∗∗
4001875
× Other 00184
4002565
Percent lean lines

× China 00003
4002495
× Thailand 00920∗∗
4002605
Table A.3 (Continued)
(1) (2)
Percent lean lines

× Vietnam 10355∗∗
4005145
× Sri Lanka 00274
4004595
× Malaysia 00912∗∗
4002515
× India 10360
4003595
× Other −00025
4005255
Factory FEs Ø Ø
Half-year FEs Ø Ø
Factories 300 300
Total observations 2,704 2,704
Notes. The table shows OLS panel fixed effects regression from FY2009 H1 to FY2014 H1. Regression
coefficients are shown with robust standard errors clustered by factory in parentheses. The outcomes are
factory labor compliance grades on a four-point scale (A = 4, B = 3, C = 2, D = 1). The two codings of our
lean measure have been interacted with country indicators to estimate treatment effects within each country
that represents at least 5% of our sample. Bangladesh, Cambodia, Egypt, Indonesia, and Turkey make up the
“Other” category. Results are plotted in Figure 3.
∗∗
p < 0001.
Table A.4 Nike Compliance Audit Scoring Rubric
Health, safety,
Grade Labor and environment
A • Isolated violations of standards that do not rise to the level of “Serious” or “Critical” issues • Fully compliant
• No more than five minor issues awaiting remediation • Demonstrates best
practices
• Considered a leader
B • Isolated violations of standards that do not rise to the level of “Serious” or “Critical” issues • Mostly compliant
• More than five minor issues awaiting remediation • Minor system failures
• Factory making progress
C • Factory not providing basic terms of employment (contracts, documented training on terms of employment, equal • Noncompliant
pay, discriminatory employment screening) • Serious system failures
• Isolated use of workers under the minimum legal age or above the minimum legal age but under the minimum age • Factory making no
of Nike’s standards progress
• Factory fails to honor a material term of a signed collective bargaining agreement
• Isolated case of not paying the legally mandated minimum wage; not providing legally required nonincome-related
benefits; or failure to provide required income-related benefits
• Isolated verbal or mental harassment or abuse
• Violation of local laws regarding the use of migrant labor
• Serious violation of hours of work standard: factory fails to provide verifiable timekeeping system to accurately
record work hours; more than 10% of employees work between 60 and 72 hours each week or work seven or more
consecutive days without a break
D • Management specifically refuses or continues to demonstrate that it is not willing to comply with Nike Standards • Noncompliant
• Any denial of access to authorized compliance inspectors • Demonstrates general
• Management provides false information (statements or documents or demonstrates coaching) disregard for Nike codes
• Factory outsources to an unapproved or unauthorized facility or issues home work to employees and standards
• Any use of bonded, indentured, or prison labor • Unwilling or unable to
• Use of force to compel illegal work hours drive important change
• Systemic use of workers under the minimum legal age for work • Deliberately misleads
• Factory denies workers freedom of association auditors
• Systematically not paying the legally mandated minimum wage or not providing legally required income-related • Audit shows critical
benefits systemic and
• Factory conducts pregnancy testing as a condition of employment widespread problems
• Systematically not providing legally required maternity leave
• A confirmed serious incident of physical or sexual abuse or systemic harassment and abuse and/or failure to timely
respond to complaint(s)
• Critical violation of hours of work standard: lack of verifiable timekeeping system results in workers not having
hours or work accurately recorded; more than 10% of employees exceed daily work hour limits, work more than
72 hours each week, or work 14 or more consecutive days without a break
Note. See Nike, Inc. (2014) for more detail on the standards that these grades refer to.
Figure A.1 (Color online) Predictors of Health, Safety, and Environment (HSE) Compliance Grades
4 Point A–D AB versus CD

Worker protection:
Occupational noise
Occupational exposure limits
PPE—respirators
Machine guarding
PPE—general
Ergonomics
General work
Heat stress
Nonionizing raditiation
Maintenance:
Confined spaces
Electrical safety
Contractor safety
Hazardous energy control
Maintenance safety
Fall protection
PMV
Health:
Canteen
Sanitation
Occupational health
Drinking water
Dormitory
Childcare
Fire and emergency action:

Fire safety
Bloodborne pathogens
Medical services—first aid
Emergency action
Chemical management:
Hazardous materials
Hazardous waste
Air emissions
AST
Waste water
Asbestos
Solid waste
PCB
UST
0 5 10 15 20 0 5 10 15 20
Variable importance from random forest
(mean decrease in Gini index)
Notes. Random forest estimates of variable importance from sample of 313 audits with detailed subscores for HSE compliance. Includes factories from the eleven
countries used in the main study. Variable importance is measured by the sum of all decreases in the Gini impurity index associated with splitting on a given
variable, normalized by the number of trees in the forest (Breiman 2002, Liaw and Wiener 2002). PPE, personal protective equipment; PMV, powered motor vehicle;
AST, above-ground storage tank; PCB, polychlorinated biphenyl; UST, underground storage tank.
References Barrientos S, Smith S (2007) Do workers benefit from ethical trade?

Angrist JD, Pischke J-S (2008) Mostly Harmless Econometrics: An Assessing codes of labour practice in global production systems.
Empiricist’ s Companion (Princeton University Press, Princeton, Third World Quart. 28(4):713–729.
NJ). Barrientos S, Gereffi G, Rossi A (2011) Economic and social upgrading
Anner M, Bair J, Blasi J (2012) Buyer power, pricing practices, and in global production networks: A new paradigm for a changing
labor outcomes in global supply chains. Institutions Program world. Internat. Labour Rev. 150(3–4):319–340.
Working Paper Series, University of Colorado Boulder, Bolder. Bartley T (2007) Institutional emergence in an era of globaliza-
Appelbaum E (2000) Manufacturing Advantage: Why High-Performance tion: The rise of transnational private regulation of labor and
Work Systems Pay Off (Cornell University Press, Ithaca, NY). environmental conditions. Amer. J. Sociol. 113(2):297–351.
Autor DH (2003) Outsourcing at will: The contribution of unjust Becker BE, Huselid MA (1998) High performance work systems
dismissal doctrine to the growth of employment outsourcing. and firm performance: A synthesis of research and managerial
J. Labor Econom. 21(1):1–42. implications. Res. Personnel Human Resource Management 16:
Bailey T, Berg P, Sandy C (2001) The effect of high-performance work 53–101.
practices on employee earnings in the steel, apparel, and medical Berg P, Appelbaum E, Bailey T, Kalleberg AL (1996) The performance
electronics and imaging industries. Indust. Labor Relations Rev. effects of modular production in the apparel industry. Indust.
54(2):525–543. Relations: A J. Econom. Soc. 35(3):356–373.
Barrientos S (2013) Corporate purchasing practices in global pro- Bhattacharya CB, Sen S, Korschun D (2008) Using corporate social
duction networks: A socially contested terrain. Geoforum 44: responsibility to win the war for talent. MIT Sloan Management
44–51. Rev. 49(2):37–44.
Bloom N, Van Reenen J (2007) Measuring and explaining management Ichniowski C, Kochan TA, Levine D, Olson C, Strauss G (1996) What
practices across firms and countries. Quart. J. Econom. 122(4): works at work: Overview and assessment. Indust. Relations: A J.
1351–1408. Econom. Soc. 35(3):299–333.
Bloom N, Eifert B, Mahajan A, McKenzie D, Roberts J (2013) Does International Labour Organization (2005) Promoting fair global-
management matter? Evidence from India. Quart. J. Econom. ization in textiles and clothing in a post-MFA environment:
128(1):1–51. Report for discussion at the tripartite meeting on promoting fair
Breiman L (2001) Random forests. Machine Learn. 45(1):5–32. globalization in textiles and clothing in a post-MFA environment.
Breiman L (2002) Manual on setting up, using, and understand- Report, International Labour Organization, Geneva.
ing random forests, v3.1. Statistics Department, University of International Labour Organization (2011) Labour administration and
California, Berkeley. labour inspection. Report V, International Labour Conference

Brown GD, O’Rourke D (2007) Lean manufacturing comes to China: 100th Sess., International Labour Organization, Geneva.
A case study of its impact on workplace health and safety. Keck ME, Sikkink K (1998) Activists Beyond Borders: Advocacy Networks
Internat. J. Occupational Environ. Health 13(3):249–257. in International Politics (Cornell University Press, Ithaca, NY).
Cappelli P, Neumark D (2001) Do “high-performance” work practices Kernaghan C (2006) U.S. Jordan Free Trade Agreement Descends Into
improve establishment-level outcomes? Indust. Labor Relations Human Trafficking and Involuntary Servitude (National Labor
Rev. 25(4):737–775. Committee, New York).
Cappelli P, Rogovsky N (1994) New work systems and skill require-
King AA, Lenox MJ (2001) Lean and green? An empirical exami-
ments. Internat. Labour Rev. 133(2):205–220.
nation of the relationship between lean production and envi-
Cheng B, Ioannou I, Serafeim G (2014) Corporate social responsibility
ronmental performance. Production Oper. Management 10(3):
and access to finance. Strategic Management J. 35(1):1–23.
244–256.
Clean Clothes Campaign (2008) Cashing in: Giant retailers, purchasing
practices, and working conditions in the garment industry. King BG, Pearce NA (2010) The contentiousness of markets: Politics,
Report, Clean Clothes Campaign, Amsterdam. social movements, and institutional change in markets. Annual
Cohen J, Sabel C (2006) Extra rempublicam nulla justitia? Philos. Rev. Sociol. 36(1):249–267.
Public Affairs 34(2):147–175. King BG, Soule SA (2007) Social movements as extra-institutional
Connor T, Dent K (2006) Offside! Labour rights and sportswear entrepreneurs: The effect of protests on stock price returns.
production in Asia. Report, Oxfam International, Oxford, UK. Admin. Sci. Quart. 52(3):413–442.
Dhanarajan S (2005) Managing ethical standards: When rhetoric Klassen RD, McLaughlin CP (1996) The impact of environmen-
meets reality. DevelopmentPractice 15(3–4):529–538. tal management on firm performance. Management Sci. 42(8):
Distelhorst G, Locke R, Pal T, Samel H (2014) Production goes 1199–1214.
global, standards stay local: Private labor regulation in the global Koh P-S, Qian C, Wang H (2014) Firm litigation risk and the insurance
electronics industry. MIT Political Science Department Research value of corporate social performance. Strategic Management J.
Paper, Massachusetts Institute of Technology, Cambridge. 35(10):1464–1482.
Dunlop JT, Weil D (1996) Diffusion and performance of modular Kytle B, Ruggie JG (2005) Corporate social responsibility as risk
production in the U.S. apparel industry. Indust. Relations: A J. management: A model for multinationals. Working paper 10,
Econom. Soc. 35(3):334–355. Corporate Social Responsibility Initiative, Cambridge, MA.
Egels-Zandén N (2007) Suppliers’ compliance with MNCs’ codes Liaw A, Wiener M (2002) Classification and regression by random-
of conduct: Behind the scenes at Chinese toy suppliers. J. Bus. Forest. R News 2(3):18–22.
Ethics 75(1):45–62. Locke R, Amengual M, Mangla A (2009) Virtue out of necessity? Com-
Flammer C (2013) Corporate social responsibility and shareholder pliance, commitment, and the improvement of labor conditions
reaction: The environmental awareness of investors. Acad. Man- in global supply chains. Politics Soc. 37(3):319–351.
agement J. 56(3):758–781. Locke R, Kochan T, Romis M, Qin F (2007a) Beyond corporate codes
Friedman E, Lee CK (2010) Remaking the world of Chinese labour: of conduct: Work organization and labour standards at Nike’s
A 30-year retrospective. British J. Indust. Relations 48(3):507–533. suppliers. Internat. Labour Rev. 146(1–2):21–40.
Gereffi G (1999) International trade and industrial upgrading in the Locke RM (2013) The Promise and Limits of Private Power: Promoting
apparel commodity chain. J. Internat. Econom. 48(1):37–70. Labor Standards in a Global Economy (Cambridge University Press,
Greene W (2004) The behaviour of the maximum likelihood estimator New York).
of limited dependent variable models in the presence of fixed Locke RM, Qin F, Brause A (2007b) Does monitoring improve
effects. Econometrics J. 7(1):98–119. labor standards? Lessons from Nike. Indust. Labor Relations Rev.
Guillén MF (1994) Models of Management: Work, Authority, and Organi-
61(1):3–31.
zation in a Comparative Perspective (University of Chicago Press,
Lund-Thomsen P, Lindgreen A (2014) Corporate social responsibility
Chicago).
in global value chains: Where are we now and where are we
Harney A (2008) The China Price: The True Cost of Chinese Competitive
going? J. Bus. Ethics 123(1):11–22.
Advantage (Penguin, New York).
Harrison A, Scorse J (2010) Multinationals and anti-sweatshop Lund-Thomsen P, Nadvi K, Chan A, Khara N, Xue H (2012) Labour in
activism. Amer. Econom. Rev. 100(1):247–273. global value chains: Work conditions in football manufacturing in
Haufler V (2001) A Public Role for the Private Sector: Industry Self- China, India and Pakistan. Development Change 43(6):1211–1237.
Regulation in a Global Economy (Carnegie Endowment for Inter- Luo X, Wang H, Raithel S, Zheng Q (2013) Corporate social perfor-
national Peace, Washington, DC). mance, analyst stock recommendations, and firm future returns.
Henisz WJ, Dorobantu S, Nartey LJ (2014) Spinning gold: The financial Strategic Management J. 36(1):123–136.
returns to stakeholder engagement. Strategic Management J. MacDuffie JP (1995) Human resource bundles and manufacturing
35(12):1727–1748. performance: Organizational logic and flexible production sys-
Hurst R, Eastham A, Gittings S, Kelkar U, Martowicz M, Muckosy tems in the world auto industry. Indust. Labor Relations Rev.
P, Reddy S, Thomas C (2011) Finding the sweet spot: Smarter 48(2):197–221.
ethical trade that delivers more for all. Report, Impactt, London. MacLean TL, Behnam M (2010) The dangers of decoupling: The
Huselid MA (1995) The impact of human resource management relationship between compliance programs, legitimacy percep-
practices on turnover, productivity, and corporate financial tions, and institutionalized misconduct. Acad. Management J.
performance. Acad. Management J. 38(3):635–672. 53(6):1499–1520.
Ichniowski C, Shaw K, Prennushi G (1997) The effects of human Milgrom P, Roberts J (1990) The economics of modern manufacturing:
resource management practices on productivity: A study of steel Technology, strategy, and organization. Amer. Econom. Rev.
finishing lines. Amer. Econom. Rev. 87(3):291–313. 80(3):511–528.
Millward Brown Optimor (2013) Top 100 most valuable global brands. Ruwanpura KN, Wrigley N (2011) The costs of compliance? Views of
Report, accessed February 22, 2016, http://www.millwardbrown Sri Lankan apparel manufacturers in times of global economic
.com/brandz/2013/top100/docs/2013_brandz_top100_report.pdf. crisis. J. Econom. Geography 11(6):1031–1049.
Monden Y (2012) Toyota production System: An Integrated Approach to Sako M (2004) Supplier development at Honda, Nissan and Toyota:
Just-In-Time, 4th ed. (Productivity Press, New York). Comparative case studies of organizational capability enhance-
Nike, Inc. (2012) Nike, Inc. FY10/11 sustainable business performance ment. Indust. Corporate Change 13(2):281–308.
Seidman G (2007) Beyond the Boycott: Labor Rights, Human Rights, and
summary. Report, accessed February 22, 2016, http://www
Transnational Activism (Russell Sage Foundation Publications,
.nikeresponsibility.com/report/uploads/files/Nike_FY10-11
New York).
_CR_report.pdf.
Shah R, Ward PT (2003) Lean manufacturing: Context, practice

Nike, Inc. (2014) Nike code leadership standard. Report, accessed bundles, and performance. J. Oper. Management 21(2):129–149.
September 10, 2014, http://www.nikeresponsibility.com/report/ Shah R, Ward PT (2007) Defining and developing measures of lean
uploads/files/NIKE_INC_Code_Leadership_Standards.pdf. production. J. Oper. Management 25(4):785–805.
Ohno T (1988) Toyota Production System: Beyond Large-Scale Production Smyth R, Zhai Q, Li X (2009) Determinants of turnover intentions
(Productivity Press, New York). among Chinese off farm migrants. Econom. Change Restructuring
O’Rourke D (2003) Outsourcing regulation: Analyzing nongovern- 42(3):189–209.
mental systems of labor standards and monitoring. Policy Studies Toffel MW, Short JL, Ouellet M (2015) Codes in context: How states,
J. 31(1):1–29. markets, and civil society shape adherence to global labor
Osterman P (2006) The wage effects of high performance work standards. Regulation & Governance 9(3):205–223.
organization in manufacturing. Indust. Labor Relations Rev. Turban DB, Greening DW (1997) Corporate social performance and
59(2):187–204. organizational attractiveness to prospective employees. Acad.
Management J. 40(3):658–672.
Oxfam International (2010) Better jobs in better supply chains.
Varian HR (2014) Big data: New tricks for econometrics. J. Econom.
Briefings for Business No. 5, Oxfam International, Oxford, UK.
Perspect. 28(2):3–27.
Pfeffer J (2007) Human resources from an organizational behavior Verité (2004) Excessive overtime in Chinese supplier factories: Causes,
perspective: Some paradoxes explained. J. Econom. Perspect. impacts, and recommendations for action. Research paper, Verité,
21(4):115–134. Amherst, MA.
Porter ME, Kramer MR (2006) Strategy and society. Harvard Bus. Rev. Vogel D (2008) Private global business regulation. Annual Rev. Political
84(12):78–92. Sci. 11:261–282.
Pruett D, Merk J, Ascoly N (2005) Looking for a quick fix: How Vogel D (2010) The private regulation of global corporate conduct:
weak social auditing is keeping workers in sweatshops. Report, Achievements and limitations. Bus. Soc. 49(1):68–87.
Clean Clothes Campaign, Amsterdam. Weaver GR, Trevino LK, Cochran PL (1999) Integrated and decoupled
Raworth K, Kidder T (2009) Mimicking “lean” in global value chains: corporate social performance: Management commitments, exter-
It’s the workers who get leaned on. Bair J, ed. Frontiers of nal pressures, and corporate ethics practices. Acad. Management
Commodity Chain Research (Stanford University Press, Redwood J. 42(5):539–552.
City, CA), 165–189. Weil D (2008) A strategic approach to labour inspection. Internat.
Labour Rev. 147(4):349–375.
Reich R (2007) Supercapitalism: The Transformation of Business, Democ-
Womack JP, Jones DT (1996) Lean Thinking: Banish Waste and Create
racy and Everyday Life (Vintage, New York).
Wealth in Your Corporation (Simon and Schuster, New York).
Riisgaard L (2009) Global value chains, labor organization and Womack JP, Jones DT, Roos D (1991) The Machine That Changed the
private social standards: Lessons from East African cut flower World: The Story of Lean Production (Harper Perennial, New York).
industries. World Development 37(2):326–340. World Trade Organization (2014) International trade statistics
Ruggie J (2008) Taking embedded liberalism global: The corporate 2014. Report, World Trade Organization, Geneva. Accessed
connection. Ruggie J, ed. Embedding Global Markets: An Enduring February 22, 2016, http://www.wto.org/english/res_e/statis_e/
Challenge (Ashgate Publishing Group, Surrey, UK), 231–238. its2014_e/its14_toc_e.htm.

002 Does Lean Improve Labor Standards Management and Social Performance in The Nike Supply Chain - Distelhorst Et Al. (2016)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

002 Does Lean Improve Labor Standards Management and Social Performance in The Nike Supply Chain - Distelhorst Et Al. (2016)

Uploaded by

Copyright:

Available Formats

This article was downloaded by: [14.139.225.

83] On: 20 April 2016, At: 03:03

Does Lean Improve Labor Standards? Management and

To cite this article:

Published online in Articles in Advance 25 Mar 2016

Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions

Copyright © 2016, INFORMS

Please scroll down for article—it is on subsequent pages

Does Lean Improve Labor Standards?

the Nike Supply Chain

1. Introduction apparel brands1 —have responded by adopting com-

choices. This pattern of improvements is consistent

in the form of longer hours, lower wages, and more

Table 1 Minimum Definitions of Nike Lean Production System (Apparel Manufacturing)

variability in production processes.

quality inspection and to communicate problems to

Nike divides its factory compliance program into

A (4) 1903 1607 001 002

FY2013 60 20 64 19 4.1. Selection Bias

Table 4 Effects of Lean Manufacturing on Compliance Scores

Dependent variable: Labor compliance HSE compliance

Imputed values: Yes No Yes No

Mean score (no lean): 2.58 2.56 2.35 2.33

Model: (1) (2) (3) (4) (5) (6) (7) (8)

Lean adoption 0029∗∗ 0042∗∗ 0008 0005

tently in the years following lean adoption, reaching a

lean certification, we observe statistically insignificant

Table 5 Binary Transformations of Dependent Variables

Dependent variable (DV): Labor compliance HSE compliance

DV coding AB = 1, CD = 0 A = 1, BCD = 0 AB = 1, CD = 0 A = 1, BCD = 0

DV mean (no lean): 0.59 0.14 0.39 0.003

Model: (1) (2) (3) (4) (5) (6) (7) (8)

Lean adoption 0015∗ 0008 0010 0000

5.1. Mechanisms for 442 audits.15 We estimate the importance of 27 audit

Figure 2 (Color online) Predictors of Labor Compliance Grades

4 Point A–D AB versus CD

Work hours under maximum

of capability building in improving labor standards in

One important limitation of this study is our limited

Dependent varaible: Labor compliance HSE compliance

Dependent varaible: Labor compliance

leanswitch t+4 000272

Table A.3 Country-Specific Effects of Lean on Labor Compliance

Percent lean lines

Table A.3 (Continued)

Percent lean lines

4 Point A–D AB versus CD

Fire and emergency action:

References Barrientos S, Smith S (2007) Do workers benefit from ethical trade?

California, Berkeley. labour inspection. Report V, International Labour Conference

Shah R, Ward PT (2003) Lean manufacturing: Context, practice

You might also like